SmiScnModel Class Reference

SmiScnModel: COIN-SMI Scenario Model Class. More...

#include <SmiScnModel.hpp>

Collaboration diagram for SmiScnModel:

Public Member Functions
SmiScenarioIndex	getNumScenarios ()
double	getScenarioProb (SmiScenarioIndex ns)
SmiScnNode *	getLeafNode (SmiScenarioIndex i)
SmiScnNode *	getRootNode ()
double	getObjectiveValue (SmiScenarioIndex ns)
double *	getColSolution (SmiScenarioIndex ns, int *length)
double *	getRowSolution (SmiScenarioIndex ns, int *length)
void	setOsiSolverHandle (OsiSolverInterface &osi)
void	setOsiSolverHandle (OsiSolverInterface *osi)
OsiSolverInterface *	getOsiSolverInterface ()
void	releaseSolver ()
	SmiScnModel ()
	~SmiScnModel ()
void	addNode (SmiScnNode *node)
Read SMPS files.
There should be three files: {name}. [core, time, stoch]. If you have different extension conventions, then you can hack the method yourself. The files can be compressed. The object that reads the files is derived from CoinMpsIO. The optional argument SmiCoreCombineRule allows user to pass in a class to override the default methods to combine core and stochastic data.
int	readSmps (const char name, SmiCoreCombineRule r=NULL)
Direct methods.
Direct methods require the user to create instances of Core data and Scenario data. Currently, the dimension of the core nodes determines the dimension of the scenario nodes, but this is something that could easily be changed.
void	processDiscreteDistributionIntoScenarios (SmiDiscreteDistribution *s, bool test=false)
	generate scenarios from discrete distribution
SmiScenarioIndex	generateScenario (SmiCoreData core, CoinPackedMatrix matrix, CoinPackedVector dclo, CoinPackedVector dcup, CoinPackedVector dobj, CoinPackedVector drlo, CoinPackedVector drup, SmiStageIndex branch, SmiScenarioIndex anc, double prob, SmiCoreCombineRule r=SmiCoreCombineReplace::Instance())
	generate scenario with ancestor/branch node identification
SmiScenarioIndex	generateScenario (SmiCoreData core, CoinPackedMatrix matrix, CoinPackedVector dclo, CoinPackedVector dcup, CoinPackedVector dobj, CoinPackedVector drlo, CoinPackedVector drup, vector< int >labels, double prob, SmiCoreCombineRule r=SmiCoreCombineReplace::Instance())
	generate scenario with labels information
loadOsiSolverData
Loads deterministic equivalent model into internal osi data structures and return handle. Note: this uses a callback class SmiCoreCombineRule to decide how to combine the core and stochastic data. The user can override the default methods when the scenario is generated (see SmiScnModel::generateScenario) or when SMPS files are processed (see SmiScnModel::readSmps).
OsiSolverInterface *	loadOsiSolverData ()
Private Attributes
CoinMessageHandler *	handler_
SmiMessage *	messages_
OsiSolverInterface *	osiStoch_
int	nrow_
int	ncol_
int	nels_
int	nels_max
double *	drlo_
double *	drup_
double *	dobj_
double *	dclo_
double *	dcup_
CoinPackedMatrix *	matrix_
double *	dels_
int *	indx_
int *	rstrt_
int	minrow_
bool	solve_synch_
double	totalProb_
SmiCoreData *	core_
SmiScenarioTree< SmiScnNode * >	smiTree_
Friends
void	SmiScnModelDiscreteUnitTest ()

Detailed Description

SmiScnModel: COIN-SMI Scenario Model Class.

Concrete class for generating scenario stochastic linear programs.

This class implements the Scenarios format of the Stochastic MPS modeling system (TODO: web pointer). Core data and Scenarios data can be passed using COIN/OSI structures, or can be read from SMPS formatted files.

Typical driver fragment looks like this

        SmiScnModel smi;
        smi.readSmps("app0110R");
        smi.setOsiSolverHandle(new OsiClpSolverInterface());
        OsiSolverInterface *osiStoch = smi.loadOsiSolverData();
        osiStoch->initialSolve();

The setOsiSolverHandle method allows the user to pass in any OSI compatible solver.

Definition at line 53 of file SmiScnModel.hpp.

Constructor & Destructor Documentation

SmiScnModel::SmiScnModel ( ) [inline]

Definition at line 179 of file SmiScnModel.hpp.

SmiScnModel::~SmiScnModel ( )

Member Function Documentation

int SmiScnModel::readSmps	(	const char *	name,
		SmiCoreCombineRule *	r = `NULL`
	)

void SmiScnModel::processDiscreteDistributionIntoScenarios	(	SmiDiscreteDistribution *	s,
		bool	test = `false`
	)

generate scenarios from discrete distribution

SmiScenarioIndex SmiScnModel::generateScenario	(	SmiCoreData *	core,
		CoinPackedMatrix *	matrix,
		CoinPackedVector *	dclo,
		CoinPackedVector *	dcup,
		CoinPackedVector *	dobj,
		CoinPackedVector *	drlo,
		CoinPackedVector *	drup,
		SmiStageIndex	branch,
		SmiScenarioIndex	anc,
		double	prob,
		SmiCoreCombineRule *	r = `SmiCoreCombineReplace::Instance()`
	)

generate scenario with ancestor/branch node identification

Core argument must be supplied. Data values combine with corresponding core values, if found, or creates them if not.

Scenario nodes need to have same dimensions as core nodes.

Data field arguments can be NULL, or empty.

branch, anc, arguments must be supplied. These identify the branching node according to the Stochastic MPS standard.

SmiScenarioIndex SmiScnModel::generateScenario	(	SmiCoreData *	core,
		CoinPackedMatrix *	matrix,
		CoinPackedVector *	dclo,
		CoinPackedVector *	dcup,
		CoinPackedVector *	dobj,
		CoinPackedVector *	drlo,
		CoinPackedVector *	drup,
		vector< int >	labels,
		double	prob,
		SmiCoreCombineRule *	r = `SmiCoreCombineReplace::Instance()`
	)